Credible Before Short

1. Why short paths are often short

On any graph of possibilities, the shortest route is easy to find. The difficulty is that a graph of possibilities can contain things that are not true. It may include a relationship between a producer and a buyer that has only been proposed. It may treat a lender as willing because it is compatible, or assume that an unknown constraint has been passed because nobody checked it. Each of these makes a path shorter. A proposed relationship becomes a step already taken, an assumed commitment removes a negotiation, and an unchecked constraint removes an obstacle.

That is why the shortest path through an unexamined graph is so often the least trustworthy. It is short because it has assumed away the hard parts. An optimizer that rewards shortness without first testing credibility will tend to favour the paths with the most invented content, since invention is the cheapest way to shorten a route. Nothing in the arithmetic would show this. The plan would look efficient and confident and would fail at the first real conversation.

The Intent Path model addresses this with a set of invariants that apply before any optimization. A named external entity must be backed by evidence actually supplied from the RSM network. "A proposed connection cannot be displayed as an existing relationship." "A candidate cannot be represented as committed, willing, qualified, or agreed without corresponding evidence." A hard constraint "cannot be hidden inside optimizer scoring." And "unknown information cannot silently become assumed truth merely to complete a plan." Each of these closes off one of the ways a path can be made shorter without being made truer.

2. What makes a path credible

The model turns these invariants into an explicit gate. "A strategy is not preferred merely because it appears short, cheap, or convenient. The optimizer SHALL first determine whether it reaches a minimum credibility condition." The first version of that gate is specified to look at seven things: hard-constraint compliance, coverage of the required goals, evidence grounding, the absence of invented relationships, the absence of false commitment assumptions, known blocking gaps, and critical unknowns.

The list is notable for what it leaves out. It contains no estimate of speed, cost or likelihood of success. Every item asks whether the plan is honest about the world: whether it respects what cannot be violated, addresses what the person actually wants, rests on evidence rather than assertion, and says openly what it does not know or cannot yet resolve. A plan with a blocking gap can still be credible, provided the gap is stated. What fails the gate is a plan that hides the gap.

The model also specifies that the gate "should remain ordinal and explainable," and it limits how much a passing result may claim. Its credibility states are insufficient, plausible, and credible for exploration. The highest of these is deliberately modest. It says a plan is sound enough to pursue, not that it will work or that anyone has agreed to it. That restraint fits the rest of the formation series. A configuration is not a Formation and a recommendation is not a commitment, and in the same way a credible strategy is not a feasible one. Possibility Is Not Agreement follows that line to its end.

3. Credibility gates; it does not rank

Credibility, then, decides eligibility rather than standing as one weight among others. The Intent Path model gives a recommended order for strategy selection that makes this concrete. First, reject plans that violate hard constraints. Second, reject plans below minimum credibility. Only then prefer plans that cover the hard requirements and plans with fewer unresolved blockers, compute the trade-off frontier across the remaining objectives, and select "a small set of meaningfully different strategies."

The order matters because credibility and convenience are not interchangeable. If they were combined into one weighted score, a sufficiently short or cheap plan could make up for an invented relationship, and the arithmetic would allow exactly the failure the gate exists to prevent. The model rejects that design by comparison: its ordering is preferable, it says, to "score = 82.7 with unclear semantics." A single number of that kind does two kinds of damage. It hides which considerations drove the result, and it suggests that a hard constraint and a soft preference can be traded against each other, when the defining feature of a hard constraint is that it cannot.

Once the gate has done its work, optimization has a legitimate place. Among credible plans it is reasonable to prefer the faster, cheaper, less burdensome or more resilient one, and the model represents these as separate dimensions: goal coverage, evidence strength, unresolved blockers, uncertainty, expected time, expected cost, risk, coordination burden and resilience. It also allows them to be ordinal "when numeric evidence is unavailable," which is honest. A system that says expected time is unknown has told the person something true. A system that reports a precise expected time on no evidence has told them something false, and made it look exact.

4. Several honest strategies instead of one hidden number

When no single plan beats the others on every dimension, the model keeps the alternatives. Its illustrative strategy set has four: buyer-backed, offering the strongest market assurance; capital-first, the fastest potential start; program-led, the lowest coordination burden; and collaborative, the greatest resilience. These are not four versions of one plan. Each makes a different bet about which dependency to resolve first and what kind of risk to carry, and which bet is right depends on what the person values, which the system should not decide for them. Only strategies the evidence actually supports should appear; the labels illustrate structure and are not a fixed menu.

Presentation is part of the design. The model asks that technical dimensions be translated into human labels grounded in the underlying measures, with each strategy shown alongside its main trade-off. A buyer-backed route gives the strongest market assurance, but buyer qualification is still required. A capital-first route may start fastest but carries more market exposure. A program-led route has a lower coordination burden, but fewer matching programs may be known. Its rule for false precision is direct: avoid "Plan score: 91.37," and prefer a statement such as evidence strong, one open blocker, coordination medium, timing confidence unknown.

Behind this is an assumption about who decides. A single score implies that the system has already done the weighing and the person's job is to accept the result. A small set of credible, structurally different strategies, each with its advantage and its cost stated, implies the reverse: the system has done the work of making the options honest and comparable, and the choice remains with the person whose outcome it is. The Seednote Should Trust Be a Score? asks the harder version of this question: what should replace the score for someone who has to decide quickly.

5. A plan is not a line

The word path has one more misleading implication: that the way forward is a single ordered sequence. Real coordination rarely works like that. Several requirements may need to be resolved in parallel, and some later steps depend on what an earlier step reveals. The Intent Path model therefore separates the plan from the path. A plan candidate is "a plausible bounded subgraph" representing one coherent way the intent might be advanced, and it may contain branches and parallel work. An intent path is "an ordered traversal through a Plan Candidate under a particular set of decision outcomes." One plan holds several paths. The model states the purpose: "This avoids pretending complex coordination problems are simple shortest paths."

The branches are defined by real decisions. The model's example is a question such as whether credible buyer demand has been established. If it has, pursue buyer-backed financing; if not, continue broader buyer discovery. Its rule is that decision points "should be used only when downstream strategy actually differs." A branch that leads to the same next step is decoration. A branch that changes what happens next is information the person needs, because it tells them which fact, once known, would change their course.

This is where credibility and dependency reasoning meet. The decision points in a good plan are usually the dependencies that matter most, the conditions which, once resolved, determine everything after them. The Commitment That Unlocks the Others describes how those dependencies can be read for leverage. A credible plan places them where the person can see them, rather than settling them silently in advance.

6. What the first version will not claim

The Intent Path model is explicit about its own limits, and they belong to the argument. Its list of things the first version will not implement includes formal Outcome Formation, formal multi-party configuration, commitment negotiation, formal feasibility probability, binding-constraint optimization, causal prediction and, notably, the "shortest credible Formation path." The specification that frames credibility as a gate declines to claim the shortest credible path at the scale of a full formation. It sets out how to compare strategies honestly at the level of a person's intent and leaves the harder multi-party problem to later formation reasoning that is not yet built.

That restraint shows the order of priorities. Credibility comes first because it is the condition under which every other measure means anything. A short path that rests on an invented relationship is not short, because the distance it skipped will reappear later. A cheap path that assumes a commitment nobody made is not cheap. Once a path is credible, with its constraints respected, its evidence grounded, its proposed relationships marked as proposed and its unknowns named, shortness becomes a reasonable thing to prefer. Before that, it is just a measure of how much the plan has assumed.

Most of this is specification. The Intent Path model describes reasoning Wellzai is designed to perform, not a system in service, and its strategy examples are illustrations. The Formation Lab on this site shows credible paths through an illustrative Formation Network. The principle can be stated independently of any implementation: a way forward deserves attention only once it is honest about the world, and the first thing a system serving regenerative outcomes should optimize is that honesty, before speed or cost.

Sources

• Wellzai Intent-to-Outcome Product and Application Architecture v1.0, §2.1 From Search to Outcome Formation.
• Wellzai Intent Intelligence and Intent Path Model v1.0, §11 Plan Candidate; §12.2 Decision Point; §16 Credibility Gate; §17 Strategy Optimization; §30 Optimization Transparency; §33 Grounding and Safety Invariants; §39 Explicit Non-Goals.